Modeling Product Affinity and Cannibalization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Retailers face challenges in accurately predicting the impact of promotional programs on customer buying decisions due to reliance on non-scientific methods and lack of objective data, leading to disappointing sales and unsold inventory, as existing models fail to effectively characterize and predict affinity and cannibalization effects between products.
Innovation Solution
A computer-implemented method using a linear relationship with a constant of proportionality between product functions, expressed through a Taylor Series expansion, to model customer responses and solve for the constant using observable data, thereby characterizing affinity or cannibalization relationships between products.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional non-scientific methods are used for promotional decisions, then implementation is simple, but prediction accuracy of customer buying decisions deteriorates
Solution Approach 1:
The patent transforms the promotional decision-making process by changing parameters from intuitive estimates to mathematically derived constants. The affinity constant (α) and cannibalization constant (β) are calculated from historical sales data using regression analysis, converting subjective promotional planning into objective parameter-based predictions that accurately forecast customer responses to promotions.
2Reliability
If complex economic models with many variables are used, then prediction comprehensiveness improves, but model reliability deteriorates due to uncertainty
Solution Approach 1:
The patent extracts the essential relationships from complex promotional dynamics by isolating two key constants: the affinity constant (α) that captures cross-product purchase relationships, and the cannibalization constant (β) that captures within-product substitution effects. This extraction simplifies the model while maintaining reliability by focusing on the most critical predictive factors.
Solution Approach 2:
The model segments promotional effects into distinct components: baseline sales, affinity-driven additional sales, and cannibalization-driven lost sales. This segmentation allows each component to be modeled separately with its own constant, improving overall prediction reliability by addressing different promotional mechanisms independently rather than as a monolithic complex model.
3Productivity
If promotional programs are expanded to maximize sales, then revenue potential increases, but cannibalization losses worsen
Solution Approach 1:
The patent implements feedback by using the calculated constants α and β to predict the net impact of promotional programs before implementation. The model forecasts both the additional sales from affinity effects and the lost sales from cannibalization effects, providing feedback that allows retailers to optimize promotional scope and pricing to maximize profit rather than simply expanding volume indiscriminately.
Data Source
AI summary
A computer system models customer response using observable data. The observable data includes transaction, product, price, and promotion. The computer system receives data observable from customer responses. A set of factors including customer traffic within a store, selecting a product, and quantity of selected product is defined as expected values, each in terms of a set of parameters related to customer buying decision. A likelihood function is defined for each of the set of factors. The parameters are solved using the observable data and associated likelihood function. The customer response model is time series of unit sales defined by a product combination of the expected value of customer traffic and the expected value of selecting a product and the expected value of quantity of selected product. A linear relationship is given between different products which includes a constant of proportionality that determines affinity and cannibalization relationships between the products.